OpenAI Launches GPT-6 Astra with SOTA Coding and Computer Use Capabilities
OpenAI has officially launched GPT-6 Astra, a new frontier model class setting state-of-the-art performance benchmarks in coding and computer use tasks. Although token costs have increased by 2.5x per token, the overall cost per completed task is significantly cheaper due to efficiency gains. GPT-6 Astra marks a significant shift in AI economics and autonomous workflow capabilities by proving that higher unit token prices can translate into cheaper task completion. Its advanced computer use capabilities bring AI closer to executing complex, multi-step digital workflows autonomously. While GPT-6 Astra delivers benchmark-defining results in coding and computer interaction, early reports note that it is less monitorable than previous models. Its pricing structure trades higher raw token rates for a lower total token requirement per finished assignment.
## BACKGROUND
Frontier AI models represent the leading edge of generative artificial intelligence, trained on massive compute infrastructure to perform broad, complex reasoning across multiple domains. Computer use capabilities allow AI models to interact with digital environments by viewing screens and performing simulated mouse and keyboard actions like a human user.